Re: [TLS] Confirming consensus: TLS1.3->TLS*

[Peter Gutmann <[e-mail]>]

Vlad Krasnov <[e-mail]> writes:

>Second: I don’t think that the changes between TLS 1.3 and TLS 1.2 are
>considered a major: just look at the difference between HTTP/2 and HTTP/1 -
>those are completely different protocols.

So are TLS 1.x and "1.3".  It'd be interesting to hear from other implementers
on this, but my secure-tunnel code consists of a high-level framework that
handles things at an abstract level, client hello, server hello, keyex, keyex-
auth, finished, and subsequent stuff, and that's the same for both TLS and SSH
(I use TLS names for consistency, but SSH does the same things under its own
names).  The bit-bagging for the two is obviously quite different, but the
high-level handling is taken from the same code.

For "1.3" I looked at what it'd take to bolt it onto the side of the other 1.x
code and it'd end up as this weird hermaphrodite mixture with huge amounts of
effort devoted to trying to track whether it's meant to be acting as 1.x or
"1.3", with the accompanying opportunity for problems if I miss something and
drop from 1.x to "1.3" or the other way round.  The easiest way to implement
it is as a new protocol, trying to pretend that 1.x and "1.3" are the same
thing just leads to an implementation nightmare when you have to keep the two
distinct.

So at least from this implementation's point of view, they're different
protocols.

Peter.